Research Overview
My research and publications can be roughly grouped into three categories: 1)
development and evaluation of network controls ; 2) network survivability and
fault management; and 3) performance modeling techniques. In the area of
network control, I have worked on the development and performance evaluation of
routing algorithms, buffer management, bandwidth allocation, queue scheduling,
flow control and call admission/access control. These problems have been studied
in the context of different network technologies such as multi-rate circuit
switched networks, frame relay service, and high speed packet switched networks
such as ATM and MPLS/IP and more recently wireless ad-hoc networks. I have been
particularly interested in the development of network controls which adapt to
changing network conditions in a manner that improves network performance and
the design of controls to support a required quality of service. I have
published a number of papers in this research area which are listed here and
have chaired or co-chaired three doctoral dissertations related to this topic .
Another focus of my research has been the problem of developing network design and traffic restoration algorithms to improve network survivability in the face of component failures or attacks. This research has been supported by grants from both industry and government agencies. The first grant was obtained from IBM while I was at Clemson University. This project started as an evaluation of IBM developed routing algorithms for their broadband networking products and evolved to a study of their network control architecture under failure conditions. Many of the recommendations of our work were incorporated into algorithms used in IBM broadband networking equipment software. Much of the work in this project was not published due to a nondisclosure proprietary agreement. However, two important publications resulted from side work on this project. The first paper, “An Analysis of the Congestion Effects of Link Failures in Wide Area Networks”, showed that a major factor on network performance after a failure is the non-stationary congestion period trigged by retransmissions after traffic restoration in fault recovery. The second paper, “An Analysis of the Timing of Traffic Restoration in Wide Area Communication Networks”, explored the use of call gapping to reduce network congestion in fault recovery in connection based packet switched networks. This initial work led to my continuing interest in network design, network survivability and traffic restoration techniques. This was the topic of an NSF Grant and a related DARPA project which focused on developing traffic restoration methods and network design tools for survivable connection based packet networks (e.g., ATM) and interconnected networks. These projects produced a number of refereed papers in the network survivability area. In particular ``Approximating Optimal Spare Capacity Allocation by Successive Survivable Routing,” which appeared in the Proceedings of IEEE Infocom 2001 (expanded version in ACM/IEEE Transactions on Networking Feb., 2005), has attracted attention and a patent for the basic algorithm in the which provides a method for near optimal real time provisioning of survivable virtual network topologies was granted in 2004.. I am continuing to work in this area and have broadened the scope to include developing network survivability mechanisms for wireless access networks and the interaction of survivability and security, which have been the focus of recent grants from NSF and NIST respectively. Related to this work, I have published papers and an invited book chapter on the survivability of wireless access networks that are among the first to appear on this topic. As wireless networking and mobile computing become ubiquitous I expected this line of research to grow in importance. A listed of papers I have published in this area are listed here.
In the area of performance modeling techniques, I have worked on developing a set of methods/algorithms for studying the time varying behavior of queueing systems either via analytical models or simulation. This work originated in my doctoral dissertation and has been extended in a number of student theses. Much of my work on network survivability and some of the models used in network control work are an application of these theoretical approaches to a specific problem. I have had funding for this work while at Pitt from MCI and a grant from IBM while at Clemson. Recently I have been working on developing performance techniques for modeling ad hoc networks. Publications related to this area are listed here.